Masters of Manipulation
Radiologists and vendors race to grasp and improve the tools of advanced visualization as imaging modalities churn out ever more information Radiology stands on the cusp of the golden age of advanced visualization, with the two most commonly used high-tech modalities, CT and MRI, increasingly reliant on 3D reconstructions and 4D analysis for examination interpretation. As advanced visualization for CT and MRI comes of age—bolstered by augmented processing power and ever-increasing transmission speeds—how can radiologists optimize the tools at their disposal for the most rapid and accurate diagnoses? Jeffrey C. Hellinger, MD, director of cardiovascular imaging and director of the 3D Medical Imaging Laboratory at The Children’s Hospital of Philadelphia, advocates an all-inclusive approach. When interpreting cardiac CT angiography (CTA) exams, Hellinger uses a combination of all tools. “The tools have strengths and weaknesses, advantages and disadvantages,” he says. “I interact freely among volume rendering, maximum-intensity projection (MIP), multiplanar reconstruction (MPR), and curved planar reconstruction.” Gary Wendt, MD, associate professor of neuroradiology at the University of Wisconsin–Madison, finds that the limitations of advanced visualization for MR angiography (MRA) are more infrastructure based. “One of the biggest problems in advanced visualization, in general, is the lack of universal accessibility,” he says. “The Web-based products are getting better, but they still have their flaws.” Each subspecialty uses a different toolset, but they all have one common goal: making interpretation as efficient as possible. As MRI resolution improves and 64-slice CT scanners gradually replace older 16-slice units, that goal is both closer and more remote than ever before. Multislice CT The evolution of CT scanners from 16 to 64 slices or more has meant an order-of-magnitude explosion in the number of images that compose a single study. While this exponential increase takes its toll on networks and IT infrastructure, it also opens new doors, allowing more detailed visualization of the heart than has ever been possible. “As the technology continues to advance, it’s more important than ever that trainees (from residents to fellows to attending radiologists from academic settings and private practice) learn to use a workstation,” Hellinger says. “You’re not going to view each dataset with each tool, but on the fly, in real time, it’s important to know instinctively whether something will be better with volume rendering, MIP, MPR, or curved planar reconstruction. All the tools have to be used in collaboration.” C. Dan Johnson, MD, chair of the department of radiology at the Mayo Clinic, Scottsdale, Ariz, specializes in CT colonography. He emphasizes the importance of learning the ins and outs of each tool. “They all have their advantages and disadvantages,” he says, “and they can all be useful in complete analysis of the colon.” Johnson recommends striking a balance between 2D and 3D visualization, stressing that both have roles to play in comprehensive interpretation. “With 2D images, you have to put the image together in your mind, so sometimes it takes a little problem solving to determine if there’s a polyp,” he notes, “but the 3D data make discrimination of stool or residual barium from polyps more difficult. The very best interpretation includes a primary 2D review of the axial images, as well as the 3D review.” Hellinger’s most common cardiovascular CTA applications include pediatric and adult congenital heart disease and congenital and acquired pediatric vascular disease. He uses volume rendering and MIP for angiographic analysis and MPR as the principal tool for diagnosis. Johnson focuses on polyp detection and finds simulated 3D flythrough and virtual dissection to be the most valuable tools in his arsenal. Both Hellinger and Johnson stress, however, that these tools must be used according to their strengths and weaknesses for the best, most efficient interpretation. “I always tell people to be flexible and apply every tool when appropriate,” Hellinger says. “If you’re not, then there’s no advantage of 3D imaging beyond the source images. We can do everything imaginable to make treatment planning efficient, but it’s really up to the end user. Radiologists have to know the principles of 3D imaging. If they approach it in an educated way, then everything’s in place for efficient interpretation and patient care.” MRI’s Midlife Crisis At 30 years old, MRI technology is reaching middle age, with higher-resolution 3T units dethroning their 1.5T predecessors. David Bluemke, MD, PhD, director of radiology and imaging sciences at the US National Institutes of Health, Bethesda, Md, notes that the modality still has its limitations. “With cardiac MRA, we’re almost always looking for narrowing or stenosis. We use fairly straightforward MIP tools, and the reason for that is it’s very quick—it can be done on the workstation or the PACS, and requires less sophisticated processing than CT. Tools are becoming available for automated detection of stenoses, but so far, they haven’t caught on,” he says. The principal issue is reliability; as he explains, “You need high-quality MRA studies, which are becoming more commonplace, but can still be difficult to get.” Wendt, who specializes in neurological MRA, echoes Bluemke. “Doing things like stenosis quantification in MRI is impossible. Advanced visualization tools like automated stenosis measurement are more applicable to CTA, where you can accurately define a lumen. With MRA, it’s more about getting a quick, easy, targeted MIP,” he says. Because much of the image’s background information is suppressed with MRA, Wendt describes the visualization and interpretation process as a quick-and-dirty approach. “It’s a technical limitation of the modality,” he says. “It’s nowhere near being as precise as CTA. MRA is all about how fast you can get at it—how fast you can generate the images.” For Bluemke, the most interesting new advanced visualization tools for MRI effectively leave behind the realm of 3D imaging to visualize flow. “These tools are new, and we don’t have a routine clinical application for them,” he says, “but 3D point tracking of flow could be used to look at the effect of flow on the vessel wall. They’re actually more 4D, because they’re time resolved. You can look at the wall’s shear stress in 3D throughout the cardiac cycle.” When it comes to incidental findings—including renal cancers, fistulae, or unsuspected aneurysms—Bluemke turns to his PACS for the right tools, such as thin-sliding MIP. “The most useful tools are those that are integrated into the PACS,” he says. “We’ve had very good software for some time, but on very specialized and expensive equipment. Integration is the key to getting the readings done quickly and accurately.” Wendt concurs. “We’re a multiple-enterprise organization,” he says. “We serve multiple hospitals and we have one PACS. We need a unified, single log-on for advanced image processing, and it needs to be tightly integrated with our PACS. Otherwise, if you try to run advanced image-processing tools in a traditional manner, it becomes challenging.” PACS Integration The need for tighter PACS integration isn’t limited to radiologists specializing in MRI. Johnson also cites workstation-based advanced visualization as a barrier to more efficient workflow. “Right now, I do interpretation on separate, high-end workstations, where the data are imported to the workstation for evaluation,” he says. “It would be very helpful if these tools were integrated into PACS. It probably wouldn’t speed up my evaluation time, but it would increase productivity if you didn’t have to leave your PACS and go to a workstation to load up a study.” Bluemke also sees the potential for more timely interpretation. “The ability to do rapid multiplanar imaging at the same time as MIP is probably what’s most important now,” he says. “Radiologists have all those images on their PACS. It’s highly desirable not to have to go to a separate workstation.” Wendt looks forward to the wide-ranging operational efficiencies that could be achieved via full multiple-enterprise PACS integration. “Then, you won’t need a whole separate deployment or a whole separate method for managing image flow,” he says. “When you’re trying to figure out how images flow through multiple organizations, you have to deal with managing flow across multiple RIS and hospital information systems. If you have to deploy not only PACS, but also multiple advanced visualization platforms, it gets impossible pretty quickly. How do you manage security, user access, and audit trails?” Describing his vendor’s process, he says that the company has had a thin client for years. Now, he adds, “It’s integrating its thick client and thin client with the integrated workstation thick client. It’s all four platforms converging onto one back end.” Hellinger looks forward to advanced visualization integration not only with PACS, but also for reporting packages and dictation systems. “I think, in the next year or so, we’ll begin to see deeper integration with all health IT,” he says, “and as we advance the technology, education on 3D imaging is more important than ever before.” Choosing Clients The most daunting issue in advanced visualization has nothing to do with the software toolset, according to Wendt. Accessibility and cost are his concerns. “With the stand-alone workstations, when you have 15 to 20 of them, every time you do an upgrade, you’re looking at a couple million bucks,” he says. “That’s pretty outrageous. They have more functionality than thin clients, but the pricing gets out of hand really fast. Likewise, I think there are vendors who oversell their thin clients. They claim to be 100% Web based, but when you hold their feet to the fire, it turns out that you do need to install software. I don’t know of any vendor that’s truly 100% Web based right now.” Bluemke takes a similarly measured view of the thin-client technology currently available. “We’re not viewing in native 3D yet,” he notes. “The images are inherently 3D when they’re generated, but we’re not routinely viewing them in 3D (Figure 1).” As advanced visualization solutions are incorporated into standard PACS and standard PC systems, however, Bluemke expects to see robust thin-client visualization systems proliferate. Hellinger is more optimistic about the direction in which the technology is heading. “First, you had thick clients; then, thin clients; and now, Web clients,” he says. “The technology will only become more robust. Ten years ago, people were really skeptical about the concept of thin clients. Now, they are the standard for imaging. I think the new standard will be Web clients—no software, just an Internet browser where you log into a server to view and interpret studies (Figure 2).” How is this possible? Wendt says that we have the Xbox 360® and Sony PlayStation 3®—among other gaming platforms—to thank for coming advances in the field. “The power that’s going into the new graphics cards is going up exponentially, and it’s driven by factors way outside the medical industry,” he notes. “If you want cutting edge, check out the Xbox. The amount of advanced graphics processing that goes on in one game is more than we do in two years.” What’s Next
Advanced visualization evolves alongside emerging clinical applications in radiology. What’s the next frontier in high-tech imaging? For Johnson, computer-assisted detection could mean a big improvement in productivity. “Computer-assisted detection has yet to be integrated in all of the commercial products yet, and I think it would be very helpful,” he says. “It may well be able to reduce the time on 3D review. If radiologists just did the 2D review and looked for other detections on 3D with computer-assisted detection, it could reduce the 3D interpretation time and improve overall exam performance.” Hellinger also looks for computer-assisted detection integration, and that’s not all. “I think there will be more of a bridge between molecular imaging and 3D imaging,” he says. “When you’re imaging at the genomic level with MRI and a radiotracer, being able to diagnose, plan treatment, and guide treatment with greater confidence will be valuable.” Meanwhile, Bluemke is most interested in the potential for strain imaging of vessel walls. “We could quantify what’s causing the strain and the change in wall strain over time. That may relate to the prognosis for aneurysm expansion and impending dissection. It will probably be done first with echocardiography and vascular ultrasound; I’d say it’s at least two to five years away,” he says. Further down the line, Hellinger expects the kind of anytime, anywhere interpretative flexibility that’s becoming commonplace in other areas of radiology. “The future is being able to access a workstation remotely anywhere, anytime—in the hospital, out of the hospital, or even from your handheld—with all these advanced tools,” he says. “There’s still a lot of innovation to be done with 3D and 4D visualization, and the success, over the years, has really been through improvements in computers and technology. It will only get better.”